First things first: Different music styles over the years compared

For this plot I have split Bring Me The Horizon’s music into three different categories, “Old music”, their first four albums, “New music”, their three most recent releases, with exception of their experimental music from Music to listen to~, which is the third “Experimental music” category. This album isn’t exactly representative of either their “new” or “old” music style, so I made it a separate category (hence “Experimental”).

It is easily visible that BMTH’s new music is much more varied than their old music. Considering that high energy + low valence = angry music, and high energy + high valence = happy music, we could conclude that their music has overall become happier. The experimental music has very low valence values, like BMTH’s old music, and is also more angry and negative. Each category seems to have outliers that fall into the ‘sad music’ area, they are labeled in the plot.

With the highest danceability (indicated by size) score being 0.6 out of a possible maximum of 1, most of their music isn’t very danceable, but their newer music generally scores higher. The speechiness seems to be higher with higher valence scores, but overall the music doesn’t seem very speechy. Looking at all their work together, the music seems very energetic and seems to have progressed from negative and angry towards a more diverse and positive style (although the valence still lies mostly below 0.5), with almost a kind of “dip” back towards negativity in their experimental album.

Popularity scores for all of BMTH’s albums

This plot shows the popularity scores for the songs on all of Bring Me The Horizon’s albums, they are in order of their release date from oldest to most recent. The bigger transparent red/pink circles show the mean popularity for each album. It is easy to see that their older music is the least popular of all and That’s The Spirit and POST HUMAN: SURVIVAL HORROR are their most popular albums. Considering the fact that their music style really started changing on Sempiternal, it is not weird to see that the popularity shoots upwards from that album on. Of course, the experimental album Music to listen to~ creates a big dip in popularity, since it is completely different from what the band usually produces and is not widely liked by fans.

Chromagrams for popular and not so popular songs

In these chromagrams we can see two very popular songs in the top row (Can You Feel My Heart and Kingslayer (feat. BABYMETAL)) and two quite unpopular songs in the bottom row (Liquor & Love Lost and A Lot Like Vegas). As discussed in the introduction, the popular songs are from BMTH’s newer music and the unpopular ones from their older music. Although Can You Feel My Heart is from Sempiternal and is still categorized as “old” music (looking at the popularity plot, it’s clearly visible where the popular songs are as outliers).

The main difference we can see between te popular and not popular songs, is that the popular ones have more distinct areas where a certain pitch is absent, as can be seen by the black areas. The high magnitude areas are also a lot clearer and more distinct than in the unpopular songs. CYFMH has a clear focus on E and C and Kinglsayer is focused on C, C#/Db and B. The unpopular songs basically have more of everything, you would be able to make out some areas of focus, but the chromagrams have a lot less contrast in them.

Chromagrams for the happiest and saddest songs in Bring Me The Horizon’s repertoire

After plotting the energy and valence of Bring Me The Horizon’s songs over the years, it gave me some insight into which songs were on the outer ends of the ‘emotional spectrum’. Because of this I decided to take a closer look at the happiest and the saddest songs in their repertoire.

Memorial is, according to it’s measurements of valence and energy, the saddest song in Bring Me The Horizon’s entire repertoire. As you can see, the song is quite repetitive and has a few layers that really stand out. Memorial is an entirely instrumental song with a continuous underlying sound and is used in the album as a transition into the next track. There are clearly audible, and in the chromagram visible, high and loud ‘screeching’ noises in the song and a few notes that are softly and repeatedly being played throughout the song. The texture of the song is very simple and also very clearly repesented in this chromagram. Interestingly, despite the fact that we saw that popular songs have a more distinct pattern, this song is not very popular. A good reason for this could be that is instrumental and belongs to the “old” music.

1x1 (feat. Nova Twins) is, according to it’s measurements of valence and energy, the happiest song in Bring Me The Horizon’s entire repertoire. As you can see, this song is all over the place. The line of C#/Db stands out a little bit, with a higher magnitude overall. The song consists of heavy drums and guitar combined with electronic music and keyboard sounds, all accompanied by the singing voices of either Bring Me The Horizon’s frontman Oliver Sykes or the featuring artist the Nova Twins. The only big visible thing that is happening here is at around 160 seconds, where the heavy instruments that support the rest of the song go quiet and we hear only a keyboard and a singing voice, which gives a bright block in the G#/Ab line.

Chromagrams for BMTH’s singles

These are chromagrams for some singles of Bring Me The Horizon, the top three being from the “old” music category and the bottom three from the “new” music category. Building on what we saw in the previous chromagrams, let’s first look at the overall contrast. It is clearly visible that the newer the music, the more contrast there is, the lighter areas are brighter and the darker areas are blacker. For Ludens, however, the chromagram is a bit brighter all over, so the contrast becomes less strong and it is more similar to Shadow Moses. This isn’t weird, as fans say this song reminds them a bit of BMTH’s old music.

Timbre changes in categories

This plot shows how the timbre components have changed throughout the different categories I divided their music into. Since the components are ordered by importance, I won’t go into all of them. For the first timbre component, representing the loudness, we can see that “old music” has a significantly longer tail than the other two categories and thus differs quite a bit in loudness. The second timbre component emphasizes the brightness, the ranges in this component are overall very high, but definitely the highest in “old music”. Although probably only the “experimental music” has a significantly lower average brightness. The third component is correlated to the flatness of the sound. You can see that the “experimental music” sounds a lot flatter than both the old and new music, which don’t have a big difference between them. The fourth component relates to sounds with a stronger attack, which clearly gets stronger going from “old” to “new” to “experimental”.The same thing goes for the fifth component. Over all of the other features, it seems that the “old music” has longer tails (a wider reach) than both of the other categories, other than that there are no gigantic differences or clear patterns there.

Cepstrograms for the most “average” songs in “old” and “new” music

These are cepstrograms for the songs that lay the closest to the average energy, valence and danceability in BMTH’s old music (Sleep With One Eye Open) and new music (Ouch). Cepstrograms show how the distribution of timbre in the music. As you can see the cepstrograms both have more intensity in the bottom, but Sleep With One Eye Open is “squished down” even more than Ouch. Both songs have a focus on the second component (brightness) and generally go up to c06. You can see that in the first component (loudness) Sleep With One Eye Open seems to be more uniform with a gradual increase at the beginning, but ouch seems to have more contrasting blocks. Ouch seems to have a flatter sound than Sleep With One Eye Open, as can be seen from looking at the third component.

Cepstrograms for BMTH’s singles: were the “average” songs representative?

These are cepstrograms for some of BMTH’s singles, with the top row belonging to the “old music” category and the bottom row to the “new music”. When comparing all these songs to the “most average songs” from both categories as seen in the previous tab, we can see that those were pretty representative for the rest of the music from those categories. The “old music” is more “squished down” into the lower timbre components and the “new music” is stronger in the higher timbre components. Even though we saw that the “average” song for “old music” was brighter (timbre component C03), that does not necessarily apply to the rest of the “old music”. As we can see in both categories, it varies from song to song. Loudness (timbre component C01) does seem to be stronger in the “new music” than in the “old music”, even though we didn’t see that clear difference between the “average songs”.

Timbre and pitch based self-similarity matrices for popular and not so popular songs

These are timbre and pitch based self-similarity matrices for some of Bring Me The Horizons most popular songs (Can You Feel My Heart and Kingslayer (feat. BABYMETAL)) and some of their least popular songs (Liquor & Love Lost and A Lot Like Vegas). Timbre wise, all but Liquor & Love Lost seem to have a quite strong pattern. Te checkerboard pattern indicates homogeniety in the overall structure of the song and shows the pattern of the different sections like verse and chorus. The checkerboard patterns on the popular songs seem to be divided into smaller sections than the pattern on A Lot Like Vegas. The greater variety in the popular songs could be a reason for their popularity. The bright criss-crossing lines indicate moments in the music that are different from everything else, Liquor & Love Lost does not seem to have a lot of those moments, which could be a reason the song is less well-liked. On the other hand, A Lot Like Vegas does seem to have quite a bit of those moments, so it is probably a combination with other factors. Pitch wise we can see some similarities with the timbre SSMs. The checkerboard pattern still seems to be divided into smaller sections for the popular music against the unpopular music. The one moment that is different from the rest in the timbre based SSM for Liquor & Love Lost doesn’t show up in it’s pitch based SSM at all.

Chordograms for BMTH’s singles

These are chordograms for some of the singles of all of Bring Me The Horizons albums in the old and new music categories (the experimental catagory doesn’t have any singles), arranged from oldest to newest (left to right, top to bottom). As you can see, there is a lot going on in all these songs. When listening to the music you can easily follow along in the chordograms. Loud percussion and distorted guitar sounds can make for bright yellow lines, like the one around 70 seconds in Shadow Moses. Considering this, it is clearly visible that Bring Me The Horizon keep using this a lot in their music over time. It is interesting to see that the screaming in Pray For Plagues leads to less intense recognition of chords, but does not necessarily create the bright yellow areas mentioned before.

When comparing the songs to see what happened to BMTH’s music over time, you could say that their music has become more varied. The older songs have a simpler pattern than the newer songs, which have a lot of changes in them.

How keys and modes change throughout Bring Me The Horizon’s music

This graph shows the keys and modes for the different categories I have divided Bring Me The Horizon’s music into. It is remarkable, that - even though their new music can generally be classified as happier and more positive, as we’ve seen earlier - we see more minor modes in that category than in their old, sadder and angrier music. Their least favourite key seems to be D# (in either mode really) and most of their songs seem to be in C, C# and G for both their old and new music. Their old music also focuses a bit more on G# major. For their experimental music, however, they focus on a key that is not very common in the rest of their repertoire, which is D. Mostly D major, while most of that music is in the minor mode. The only thing that can not be derived from Spotify’s information, is whether they use sharps (#) or flats (♭).

Tempo analysis for Bring Me The Horizon’s repertoire

This plot displays the tempo in beats per minute for Bring Me The Horizon’s entire repertoire, the albums are sorted by their release date and each dot represents a song. The bigger red/pink circles represent the median for the tempo for each album. It is interesting to see that their old music has the lowest tempo and it starts climbing up when their music starts to drift away from the “metalcore” genre. In their newer music the mean line almost seems to be coherent with the means of popularity for these albums.

Loudness for Bring Me The Horizon’s repertoire

This plot shows the loudness per track for all of BMTH’s albums, from oldest-newest from left to right, coloured by category (I have scaled the Y-axis to remove one song from There is a Hel…, as it had a loudness of -22 and made the rest of the plot hard to read). The larger red/pink circles show the median of the loudness per album. As we have seen in the timbre analysis, the “new music” is slightly stronger in the timbre component that indicates loudness and that is visible in this plot. This plot also shows, as a lot of other parts of the analysis have, that the “experimental music” has very different scores from the rest of BMTH’s music.

Classification

Here I have trained a classifier to distinguish between the three different categories I divided Bring Me The Horizon’s music into. After trying different combinations, I figured out that energy, valence, duration and all of the timbre components are the most useful features for computational recognition of this corpus. The keys generally confused the classifier the most and duration seemed to influence the experimental category the most (significantly fewer songs were classified correctly). As you can see in the mosaic plot, most of the music is classified in the right category. It is important to mention that for both old and new 30 songs were used, but there are only eight experimental songs. Considering this, it’s quite remarkable that five out of those eight are classified correctly here.

The precision rates are very high for all of the categories, 83% for experimental music, 82% for new music and even 96% for old music. The categories I divided the music into are thus very distinguishable from each other, especially when filtering out the right elements (even without doing this the scores are still relatively high).

add energy/valence/duration plots for clarification?